1. Field of the Invention
This invention generally relates to audio encoding and decoding systems (hereinafter, simply referred to as audio encoding-decoding systems) which perform encoding and decoding with respect to audio signals transmitted on communication lines. Particularly, this invetion relates to audio encoding systems which perform compressive encoding on audio signals by performing vector quantization, using code books, on residual signals corresponding to results of analysis of linear predictive coding made on audio signals.
2. Prior Art
Conventionally, the encoding method of so-called `CELP` type (where `CELP` is an abbreviation for `Code-Excited Linear Prediction`) is known as the compressive encoding method which is capable of performing compressive encoding (or compressive coding) on audio signals with a low bit rate and with high quality. According to the encoding method of the CELP type, vector quantization is performed using a code book with respect to residual components which correspond to results of the analysis of the linear predictive coding (hereinafter, simply referred to as `LPC analysis`). Herein, the LPC analysis is effected on audio signals which are extracted from waveforms by certain intervals so as to calculate LPC coefficients. Quantization is performed on the LPC coefficients. In addition, the method calculates residual signals based on the LPC coefficients to produce gains which are then subjected to quantization. Using the gains, the residual signals are subjected to normalization. Thereafter, the method uses the technique of so-called MDCT (where `MDCT` is an abbreviation for `Modified Discrete Cosine Transformation`), for example, to convert the residual signals of time series into signals of frequency ranges. Those signals are divided to match with appropriate sub-frames and are then subjected to vector quantiation using the code book. Thereafter, the method performs composition on `quantized` LPC coefficients, gains and vector quantization indexes to produce bit streams of compressive coding (simply, referred to as `compressed` bit streams). Thus, a series of operations of the compressive coding are completed. Next, the decoding method performs decomposition on the compressed bit streams to reproduce the LPC coefficients, gains and vector quantization indexes, which are then subjected to reverse quantization and composition to produce decoded signals.
Among the known encoding methods of the CELP type, there is provided a method using conjugate structure code books which improve durability of transmission errors in communications. An example of this method is shown by the paper entitled "8 kbit/s audio encoding using conjugate structure CELP", provided by the Japanese people of the names of Kataoka, Moriya and Hayashi, which is written on pp 273 of the lecture paper collection of Japanese Acoustics Society, dated October of 1992. According to this method, vector quantization is performed using a pair of code books which are in conjugate relationship with each other. Thus, this method is capable of providing an advantage which copes with an error event that a transmission error occurs in an index of one side of a communication line, as follows:
Even in the above error event, it is possible to reduce a degree of influence due to the transmission error on the basis of an index of another side of the communication line.
In addition, the conventional technology provides another type of the method which uses two-stage vector code books to further improve quality of reproduction of original sound. According to this method, a first vector is selected to be an optimum one for a main code book; then, a second vector is selected from a supplementary code book. Herein, the second vector is combined together with the first vector to provide a "combined" vector. So, the second vector is selected from the supplementary code book in such a way that the combined vector approaches a target vector as close as possible.
The conventional audio encoding-decoding system described above has a variety of advantages as follows:
The conjugate structure code books are used to raise redundancy of transmitting information, so it is possible to improve durability of the system against transmission errors. Therefore, it is possible to perform transmission of information with high quality even in a poor environment of communications. Further, it is possible to perform transmission of information with high quality by two-stage coding.
However, the conventional system suffers from a problem that a bit rate is increased to damage real-time performance of communications. In the conventional system, a bit rate of transmission is directly determined by a coded mode which is set in advance. If transmission of audio signals is performed in real time under a specific environment, such as an environment of the Internet, where communication bands vary in real time in response to a degree of congestion of communication lines, the conventional system has a difficulty to enable transmission of information without pauses by the preset bit rate when the lines are congested. Such a situation damages real-time performance of transmission.
Moreover, the conventional system has another kind of problem with respect to the recording of audio information to recording media. That is, to raise a sound quality of recording, an amount of audio information which can be accumulated in the recording media should be reduced. In general, a sound quality of reproduction depends upon an amount of information secured. For this reason, it is difficult to directly set an amount of coded information to be recorded in the recording media.